TY - GEN
T1 - Influence of the Applied Outlier Detection Methods on the Quality of Classification
AU - Moska, Błażej
AU - Kostrzewa, Daniel
AU - Brzeski, Robert
N1 - Publisher Copyright:
© 2020, Springer Nature Switzerland AG.
PY - 2020
Y1 - 2020
N2 - This paper presents a comparison of a few chosen outlier detection methods and test quality of classification, both before and after the procedure of removing outliers. Using a few selected methods of outlier detection on several selected data sets, the process of elimination of atypical data was carried out. Atypical data may be of various nature. It can be noise or can be incorrect data. However, they can also be correct data, which for some reason are different from typical data. The removal of non-typical data may have a different effect on the classification quality. It may be dependent on the used method of removing unusual data but also on the nature of used data. Therefore, the classification process was carried out on the original data as well as with the outliers removed. The obtained results were compared and discussed.
AB - This paper presents a comparison of a few chosen outlier detection methods and test quality of classification, both before and after the procedure of removing outliers. Using a few selected methods of outlier detection on several selected data sets, the process of elimination of atypical data was carried out. Atypical data may be of various nature. It can be noise or can be incorrect data. However, they can also be correct data, which for some reason are different from typical data. The removal of non-typical data may have a different effect on the classification quality. It may be dependent on the used method of removing unusual data but also on the nature of used data. Therefore, the classification process was carried out on the original data as well as with the outliers removed. The obtained results were compared and discussed.
KW - Angle based outlier detection
KW - Classification
KW - Data analysis
KW - Distance-based method
KW - Interquartile method
KW - Local outlier factor
KW - Outlier detection method
UR - https://www.scopus.com/pages/publications/85075872210
U2 - 10.1007/978-3-030-31964-9_8
DO - 10.1007/978-3-030-31964-9_8
M3 - Conference contribution
AN - SCOPUS:85075872210
SN - 9783030319632
T3 - Advances in Intelligent Systems and Computing
SP - 77
EP - 88
BT - Man-Machine Interactions 6 - 6th International Conference on Man-Machine Interactions, ICMMI 2019
A2 - Gruca, Aleksandra
A2 - Deorowicz, Sebastian
A2 - Harezlak, Katarzyna
A2 - Piotrowska, Agnieszka
A2 - Czachórski, Tadeusz
PB - Springer
T2 - 6th International Conference on Man-Machine Interactions, ICMMI 2019
Y2 - 2 October 2019 through 3 October 2019
ER -